This implies that each hospital contributes to the estimated effect proportional to its yearly number of hospital outpatient visits. Similarly, variance estimates were obtained by weighted estimation, so that confidence intervals for change in presence of information reflect the variation within and between hospitals. Whenever estimated 95%confidence intervals did not include zero, it was interpreted as a significant change in presence of information. In all analyses, the statistical software R8 was applied. The package survey was used for performing weighted estimation and confidence intervals for change in presence of information.